When Should We Biopsy? A Risk Factor-Based Predictive Model for EIN and Endometrial Cancer.

BACKGROUND

The incidence of endometrial cancer (EC) is rising globally across all age groups. Endometrial intraepithelial neoplasia (EIN) is a premalignant lesion that may progress to EC if untreated. A clinical model is needed to efficiently identify women requiring prompt evaluation while avoiding unnecessary invasive procedures. Obesity is a major risk factor, but whether Asian women require a lower body mass index (BMI) cutoff than the World Health Organization (WHO) definition remains debated. This study aimed to develop a multivariable risk prediction model to guide biopsy decisions and determine an appropriate BMI cutoff for predicting EIN/EC risk among Asian women.

METHODS

This study retrospectively reviewed 1192 women aged ≥18 years who underwent hysteroscopy between 2010 and 2023 at a tertiary hospital. Candidate predictors included patient age, parity, BMI, postmenopausal status, symptom of abnormal uterine bleeding (AUB), diabetes mellitus, hypertension, polycystic ovary syndrome (PCOS), use of oral contraceptives, intrauterine devices, or menopausal hormone therapy, tamoxifen treatment, presence of multiple polyps, and endometrial thickness (EMT) measured by transvaginal ultrasonography. Multivariable logistic regression with stepwise selection identified independent predictors, and model stability and calibration were assessed using 1000 bootstrap resamples.

RESULTS

EIN/EC was diagnosed in 55 patients (4.6%). Six independent predictors were identified: postmenopausal status (adjusted odds ratio [aOR] 5.93, 95% CI 2.92-12.04), AUB (aOR 4.07, 1.51-10.97), multiple polyps (aOR 2.49, 1.33-4.66), PCOS (aOR 2.37, 1.08-5.22), BMI (aOR 1.13 per kg/m2; 1.84 per +5 kg/m2), and EMT (aOR 1.07 per mm, 1.02-1.11). When using categorical cutoffs, Obese II (BMI ≥ 30 kg/m2) and markedly increased EMT (≥20 mm) remained significant. Predicted probabilities ranged from 0.3% with no risk factors to 90.9% with all six risk factors present. The final model demonstrated good discrimination (AUC 0.79, 95% CI 0.73-0.86) and excellent calibration on bootstrap validation (mean absolute error 0.005).

CONCLUSIONS

This six-factor clinical model stratifies individual EIN/EC risk using readily available variables and may guide timely, risk-based biopsy decisions by identifying high-risk patients while minimizing unnecessary procedures in low-risk cases. BMI ≥ 30 kg/m2 (WHO obesity threshold) was confirmed as a meaningful cutoff, but external validation is warranted to confirm its generalizability and clinical applicability.